Shipping, customs and technical documents
Bills of lading, customs declarations, drawings and datasheets — documents built around numbering standards and tables rather than prose.
The documents in this cluster share a property that most business paperwork does not have: they were designed to be checked. A container number carries a check digit. A ship carries a seven-digit number whose last figure is a weighted sum of the first six. A manifest has a total that must equal the sum of its lines, a customs entry has a duty that should follow from a value and a rate, and a drawing has a revision letter in its title block that must match the top row of its revision table. None of those checks needs a human, a second source, or a confidence score to run.
That changes what these pages are about. The interesting question is almost never how to prompt a model to read a form; it is what the form guarantees, what it merely asserts, and which of its fields look like one number but are three. So each page here starts from the standard — ISO 6346, the Harmonized System, Incoterms, ISO 7200, ASME Y14.5, the GHS sixteen-section safety data sheet — and works outward to the failures that survive a clean-looking extraction.
Extracting Container and Weight Details From a Bill of Lading
The container number on a bill of lading can be verified arithmetically, and the word “weight” on the same document refers to at least three different numbers.
10 min read
Extracting Tariff Codes From a Customs Declaration Form
Only the first six digits of a tariff code mean the same thing everywhere; the digits after them, and any second code from a national chapter, are where the duty is actually decided.
10 min read
Extracting Declared Value and Duty Fields From a Customs Form
The declared value on a customs form depends on the delivery term printed beside it, and the duty figure gives you an arithmetic check on both.
10 min read
Extracting Container and Cargo Details From a Shipping Manifest
A manifest states its own totals, which makes a skipped or double-counted line detectable — provided you know which rows are data and which are carried-forward subtotals.
10 min read
Extracting Voyage and Vessel Information From a Shipping Manifest
The vessel name is mutable and unverifiable; the IMO number is neither, and its check digit takes six multiplications to verify.
9 min read
Extracting Title Block Data From an Engineering Drawing
The title block is standardised in position and largely standardised in content, but ISO and ASME drawings disagree on the field set — and on a symbol that has no text at all.
10 min read
Extracting Revision History From an Engineering Drawing
Revision letters are not the alphabet, so the obvious gap check raises false alarms — here is the sequence ASME Y14.35 actually defines and the cross-check that finds a genuinely missing revision.
10 min read
Extracting GD&T Callouts From a Mechanical Drawing
A feature control frame is meaningless as loose text: it needs the symbol resolved, the datums kept in order, and the feature it points at identified.
10 min read
Extracting Electrical Ratings From a Component Datasheet
A datasheet states three different kinds of voltage limit in three near-identical tables, and taking a number from the wrong one gives you a value the part cannot survive.
10 min read
Extracting Pinout Tables From a Component Datasheet
One datasheet describes several packages, so a pin number means nothing without the package it belongs to — and the pin count in the package name is often wrong.
10 min read
Extracting Hazard Statements From a Safety Data Sheet
H-codes are language-independent and precise, but four of them are published with a blank the supplier fills in — so the code alone loses information the printed text carries.
10 min read
Extracting First-Aid and Handling Sections From a Safety Data Sheet
The sixteen sections of a safety data sheet are fixed in number and order, which makes segmentation a solved problem — provided you anchor on the numbers and not the titles.
10 min read
Other topics
- LLM fundamentals & architecture
- Tokens, tokenization & context windows
- Prompt engineering
- Reasoning models & test-time compute
- Multimodal AI: vision, audio, video
- RAG & retrieval
- Embeddings & vector search
- AI agents & tool use
- Structured output & function calling
- Fine-tuning & post-training
- Local inference errors, string by string
- Running local models day to day
- Testing code that calls an LLM
- Snapshot and property testing for model output
- Regression suites for prompts
- Eval gates in CI
- Flaky tests against a model
- Determinism and the cost of testing
- Contract and streaming tests
- Testing tool calls and retrieval
- Inference, serving & latency
- Rolling out a prompt change
- Testing AI systems in practice
- Forecasting a time series
- Machine learning on tabular data
- Geospatial data and models
- Understanding audio that is not speech
- Understanding video
- Core computer vision tasks
- Machine learning on graphs
- Point clouds and 3D
- Evaluation, benchmarks & LLM-as-judge
- Sensor and IoT data
- Logs and event streams
- Models over biological sequences
- Machine learning on molecules
- Embedding and searching code
- Extracting invoices and purchase orders
- Receipts, statements and tax forms
- Insurance policies and contracts
- Deeds, court filings and patents
- Extracting from medical records
- Observability & LLMOps
- CVs, certificates and identity documents
- Meetings, email, chat and filled-in forms
- Building an extraction pipeline
- Business, property and inspection documents
- Contract clauses and insurance claims
- Regulated and compliance documents
- Consumer, travel and closing documents
- Mapping one chat API onto another
- SDK and framework migrations
- Hallucination & failure modes
- Re-embedding and model deprecation
- Cutting over between providers
- Parity gaps, shims and legacy endpoints
- Moving between model versions
- Migrating vector stores and caches
- Mapping capabilities and parameters
- Migrating pipelines and agents
- Contracts, runbooks and rollback
- Auditing a codebase before a cutover
- Compliance and fine-tune migration
- LLM cost engineering
- Routing, cost tracking and multi-tenancy
- What a migration does to your prompts
- AI security & prompt injection
- Privacy, compliance & data residency
- AI governance, policy & society
- Building reliable AI applications
- AI hardware, GPUs & compute
- Open-weight models & local inference
- AI for developers & coding agents
- AI in industry: vertical playbooks
- AGI, superintelligence, alignment & the long future
- Machine learning foundations
- NLP fundamentals & classical tasks
- Data engineering for AI
- Synthetic data & dataset curation
- AI product design & UX
- Search, ranking & recommendation
- Enterprise adoption & change management
- AI careers, skills & teams
- Reading AI research
- AI in science & discovery
- Robotics & embodied AI
- AI economics, markets & business models
- AI myths, hype & media literacy
- Context engineering
- Shipping AI features: patterns & anti-patterns
- Build it: end-to-end AI tutorials
- Python for AI: hands-on recipes
- TypeScript, React and the web
- Frameworks and SDKs
- Errors and troubleshooting
- AI facts, numbers and statistics
- The history of AI
- The maths behind AI
- Architectures beyond the transformer
- Reinforcement learning
- Diffusion and generative media
- Speech, audio and voice engineering
- Benchmarks, one at a time
- AI search visibility
- Infrastructure and operations
- Databases and storage for AI
- Knowledge graphs and structured knowledge
- Classical ML in production
- Regulation, jurisdiction by jurisdiction
- Prompt recipes and pattern library
- AI for people who do not write code
- Writing, media and creative work
- Edge and on-device AI
- Interpretability and model internals
- Field notes
- OpenAI model behaviour
- Claude model behaviour
- Gemini model behaviour
- Llama model behaviour
- Mistral model behaviour
- Qwen model behaviour
- DeepSeek model behaviour
- Cohere model behaviour
- Grok model behaviour
- Small model behaviour
- Hybrid model architectures
- Token cost by language and script
- Transliteration, romanization and script handling
- Locale-correct output
- Multilingual generation quality
- Multilingual pipelines
- The EU AI Act, article by article
- AI under the GDPR and EU data law
- US AI regulation, state and sector
- International AI governance and standards
- AI litigation and enforcement
- Running AI workloads on AWS
- Running AI workloads on Google Cloud
- Running AI workloads on Azure
- AI at the edge: Workers, Vercel and Netlify
- Serving models on Kubernetes
- Operating AI infrastructure
- Quantization formats and what they cost
- llama.cpp, flag by flag
- Ollama and the desktop local-model runtimes
- Local models on Apple Silicon
- Hardware for local inference
- Running speech and embedding models locally
- Model files, adapters and conversion
- VRAM arithmetic for local models